AWS Knowledge Hub

English

Tiếng Việt

English

Tiếng Việt

Appearance

EC2 Cost Optimization ​

Summary ​

  • EC2 can dominate your AWS bill; costs come from compute (instances), storage (EBS, snapshots), data transfer, and supporting services (NAT, load balancers, etc.).
  • AWS offers several pricing models (On‑Demand, Reserved Instances, Spot, Savings Plans) so you can align cost with workload predictability and fault tolerance.
  • Effective optimization combines right‑sizing, selecting the right pricing model, and good lifecycle/data‑transfer management (Auto Scaling, stopping/terminating, and minimizing cross‑Region/Internet traffic).

Cost optimization strategy mindmap ​

mermaid
mindmap
  root((EC2 Cost Optimization))
    Right-Sizing
      Monitor CPU/Mem/IO
      Resize instances
      Optimize storage
    Pricing Models
      On-Demand
      Reserved Instances
      Spot
      Savings Plans
    Lifecycle
      Auto Scaling
      Stop/Terminate unused
      Cleanup EBS/Snapshots/EIPs
    Data Transfer
      Same-Region design
      VPC Endpoints
      CloudFront & caching

Best Practices ​

  • Right‑size before you commit: collect 2–4 weeks of utilization data (CPU, memory, network, disk) and shrink or change instance types and EBS volumes before buying RIs/Savings Plans.
  • Layer pricing models:
    • Predictable baseline → Reserved Instances/Savings Plans.
    • Batch, CI/CD, ML experiments → Spot Instances.
    • Short‑lived or experimental → On‑Demand.
  • Use Auto Scaling instead of static fleets; scale‑in during low‑demand hours to avoid paying for unused capacity.
  • Optimize storage costs by preferring gp3, using io1/io2 only when required, cleaning up unattached EBS volumes, and pruning old snapshots with lifecycle policies.
  • Reduce data transfer cost by keeping traffic in the same Region/AZ, using VPC endpoints for S3/DynamoDB, and leveraging CloudFront and caching to reduce internet egress.
  • Implement a tagging strategy (Environment, Project, Owner, CostCenter) and use AWS Budgets/Cost Explorer/Cost Anomaly Detection to track and alert on spending.

Exam Notes ​

  • Understand the trade‑offs of each pricing model:
    • On‑Demand → maximum flexibility, highest hourly cost.
    • Reserved Instances / Savings Plans → up to ~72% savings for steady‑state workloads.
    • Spot → up to ~90% savings but can be interrupted (2‑minute warning).
  • In exam questions, steady long‑running workloads usually point to Reserved/Savings Plans, while batch/CI pipelines suggest Spot and dev/test suggest On‑Demand or Spot.
  • Remember that stopped instances still incur EBS charges, and idle Elastic IPs incur charges as well; both are common sources of avoidable cost.
  • For cost management, AWS promotes Cost Explorer, AWS Budgets, Cost Allocation Tags, and Cost Anomaly Detection as first‑class tools.

AWS documentation ​